Effectiveness and safety of rivaroxaban vs. apixaban in patients with atrial fibrillation and peripheral artery disease

Authors: Loubna Dari, Sarah Beradid, Joël Constans, Antoine Pariente, Christel Renoux

Abstract

Aims

To assess whether rivaroxaban is associated with a decreased risk of major adverse limb events (MALE), stroke, systemic embolism (SE), and major bleeding (MB) among patients with non-valvular atrial fibrillation (NVAF) and peripheral artery disease (PAD), compared with apixaban.

Methods and results

We conducted a population-based cohort study using the UK Clinical Practice Research Datalink. Patients aged ≥45 years with incident NVAF and PAD who initiated rivaroxaban or apixaban between 2013 and 2021 were included. Primary effectiveness outcomes were MALE, and a composite of ischaemic stroke, transient ischaemic attack (TIA), or SE. The primary safety outcome was MB. The risk of major cardiovascular events (MACE) was assessed as a secondary outcome. Confounding was addressed using propensity score fine stratification and weighting. Weighted Cox proportional hazards models estimated hazard ratios (HRs) with 95% confidence intervals (CIs). The cohort included 6170 new users of rivaroxaban and 9990 new users of apixaban (44% female; mean [SD] age 78.5 [9.2] years). Incidence rates were similar for MALE (6.7 vs. 5.6/1000 person-years; adjusted HR (aHR): 1.20; 95% CI 0.87–1.65), stroke/TIA/SE (24.5 vs. 21.3/1000 person-years; aHR: 1.15; 95% CI 0.97–1.36), and MACE (40.1 vs. 35.9 per 1000 person-years; aHR 1.10: 95% CI 0.94–1.28). Major bleeding rates were higher with rivaroxaban (46.1 vs. 29.8/1000 person-years; aHR: 1.55; 95% CI 1.36–1.77).

Conclusion

In patients with NVAF and PAD, rivaroxaban was associated with a similar risk of MALE and stroke/TIA/SE, but a higher risk of MB compared with apixaban. These findings support apixaban as a potentially safer anticoagulant in this high-risk population.

Introduction

Peripheral artery disease (PAD) is a significant risk factor for stroke and cardiovascular death in patients with atrial fibrillation (AF).^1-3^ Patients with both PAD and AF also experience worse limb outcomes, including major limb events (MALE).^4^ Typically, anticoagulation for stroke prevention in non-valvular AF (NVAF) is recommended in patients with a CHA2DS2-VASc score ≥ 2 with direct oral anticoagulants (DOACs) as a first line treatment.^2,5^ While international guidelines do not provide guidance on the choice of DOAC, observational studies suggest a higher incidence of major bleeding (MB) with rivaroxaban compared to apixaban, despite similar effectiveness in preventing stroke and systemic embolism (SE).^6-9^

In patients with PAD alone however, recent trials have demonstrated the efficacy of low-dose rivaroxaban combined with aspirin compared with aspirin alone for the prevention of major cardiac events (MACE) and MALE.^10,11^ Nonetheless, the benefit of rivaroxaban was only assessed in patients without NVAF since patients with an indication for standard-dose anticoagulants were excluded. Only one small cohort study has compared the effectiveness and safety of rivaroxaban vs. apixaban in patients with NVAF and atherosclerotic disease (coronary artery disease or PAD).^12^ The risk of stroke and SE as well as major bleeding was higher with rivaroxaban compared to apixaban. However, <40% of the patients had PAD and no analysis was conducted separately in this population. Also, the study did not evaluate the comparative incidence of MALE, a major outcome in patients with PAD. Thus, the objective was to assess the effectiveness and safety of rivaroxaban compared with apixaban in patients with NVAF and PAD, in a real-world setting.

Methods

Data source

We conducted a population-based cohort study using the United Kingdom Clinical Practice Research Datalink (CPRD GOLD and Aurum databases).^13,14^ The CPRD is a large primary care database containing anonymized electronic medical records for over 60 million patients across more than 2000 practices in the UK.^13,14^ The CPRD has been shown to be representative of the general UK population in terms of sex, age, and ethnicity. The data collected encompasses demographic information, lifestyle factors, medical diagnoses, laboratory results, prescriptions issued by general practitioners, and referrals to medical specialists or hospitals. Medical diagnoses and procedures are recorded using the Read and SNOMED-CT coding systems.^15^ Prescriptions issued by general practitioners are automatically logged using a drug dictionary based on the British National Formulary. Data quality controls are conducted regularly, and many studies have shown the validity and high quality of the recorded data.^14,15^ The CPRD data was linked with the Hospital Episode Statistics (HES) Admitted Patient Care dataset and the Office for National Statistics (ONS).^16^ The HES includes information on admission and discharge dates, primary diagnoses coded using the International Classification of Diseases, Tenth Revision (ICD-10), and procedures classified according to the UK Office of Population, Census and Surveys classification 4.6 framework. The ONS includes date and causes of death. The study protocol was approved by the CPRD Research Data Governance (No. 24_003782) and the Research Ethics Board of the Jewish General Hospital, Montreal, Canada.

Study population

We formed a base cohort of all patients aged ≥45 years with a first diagnosis of AF between 1 January 2010 and 29 March 2021 (end of ONS and HES data availability) and a diagnosis of PAD after 45 years old, before or on the date of AF diagnosis. We excluded patients with less than one year of registration in CPRD before their AF diagnosis, those with a prior AF diagnosis to only include incident AF, and those prescribed an oral anticoagulant in the year before AF diagnosis. Within this base cohort, we identified all patients with a first prescription for rivaroxaban or apixaban between 1 January 2013 (apixaban being the latest approved at the end of 2012) and 29 March 2021. Cohort entry was defined as the date of the first prescription for rivaroxaban or apixaban following AF diagnosis. We excluded patients with prior use of rivaroxaban or apixaban at any time before cohort entry, as well as those initiating two oral anticoagulants at cohort entry. To ensure the inclusion of only patients with NVAF, individuals with a history of valvular surgery or rheumatic valvular disease at any time before cohort entry were excluded. We also excluded patients with hyperthyroidism, dialysis or end-stage kidney disease in the 90 days before cohort entry and patients who had a diagnosis of venous thromboembolism or underwent hip or knee joint surgery in the 30 days prior cohort entry.

All patients were followed until the occurrence of the specific outcome under investigation, discontinuation or switch to another anticoagulant, death from any cause, end of registration with the general practice, or the end of the study period (29 March 2021), whichever occurred first.

Exposure definition

We used an as-treated exposure definition where patients were considered exposed to rivaroxaban or apixaban from the date of the first prescription and censored at the date of treatment discontinuation (end of the grace period), switch from rivaroxaban to apixaban or vice versa, to another DOAC, or to VKAs. Patients were considered continuously exposed if the duration of one prescription overlapped with the date of the subsequent prescription, with a 30-day grace period in the event of non-overlapping prescriptions.

Outcomes

The primary effectiveness outcomes (i) MALE defined as a hospitalization for an ALI or amputation above the ankle and (ii) a composite of hospitalization with an incident ischaemic stroke, TIA, or SE. The primary safety outcome was MB defined as any bleeding requiring hospitalization or resulting in death. As a secondary outcome, we investigated MACE, a composite of hospitalization for myocardial infarction, stroke/TIA/SE, or cardiovascular death. We also considered a broader definition of MALE including any lower limb arterial revascularization and amputations below the ankle. Finally, we assessed a composite outcome of MACE and MALE. All outcomes were defined using relevant ICD-10 codes (primary position in non-elective hospitalization) and procedures codes in HES or in ONS (primary cause of death) (see Supplementary material online, Table S1).

Covariates

The covariates included demographic characteristics [age (modelled using cubic splines), sex, ethnicity], calendar year of cohort entry, most recent measures of alcohol abuse, body mass index and smoking status within five years before cohort entry, and the following comorbidities, measured at any time before cohort hypertension, diabetes (all types), congestive heart failure, myocardial infarction, coronary artery disease, pacemaker or implantable cardioverter-defibrillator, prior ischaemic stroke/TIA, duration of PAD prior to cohort entry, SE (including ALI), coronary artery bypass surgery or percutaneous coronary intervention, lower limb revascularization, lower limb amputations, clinical presentation of PAD [claudication or chronic limb threatening ischaemia such as trophic disorders (ischaemic ulcers/gangrene) and ischaemic rest pain], bleeding events, venous thromboembolism, anaemia, thrombopaenia, abdominal or popliteal aneurysm, cancer (other than non-melanoma skin cancer), chronic obstructive pulmonary disease, chronic kidney disease, and liver disease. Time from NVAF diagnosis to rivaroxaban/apixaban initiation was also included. We considered the following medications measured in the year before cohort antiplatelet agents, antihyperglycaemic medications, antihypertensive drugs, lipid-lowering drugs, antiarrhythmics, non-steroidal anti-inflammatory drugs, antidepressants, antipsychotics, antiepileptic drugs, proton pump inhibitors and H2 blockers, and hormonal replacement therapy. Oral anticoagulants were also included for patients with previous use before cohort entry. Finally, we included the number of hospitalizations in the year before cohort entry as surrogate marker for overall health.

Statistical analysis

We used propensity score (PS)-based fine stratification and weighting to control for potential confounding.^17^ The PS (probability of rivaroxaban initiation) was estimated using logistic regression including all covariates described above, separately for patients with and without prior OAC use. Following PS estimation, patients in the non-overlapping regions of the PS distributions were excluded. We created 100 strata based on the PS distribution of the treated group (rivaroxaban) and patients in the apixaban group were weighted proportionally to the number of rivaroxaban patients in the corresponding stratum. We described the baseline characteristics of each exposure group before and after weighting, by prior OAC use strata. A standardized mean difference lower than 10% was indicative of good covariate balance. We also used stabilized inverse probability of censoring weights to account for potential informative censoring. Specifically, we separately estimated the conditional probabilities of remaining uncensored, considering treatment termination or switching, administrative censoring, and death. In these three models, we used baseline covariates previously described, as well as time-updated covariates. We estimated the incidence rates with 95% confidence intervals (CIs) of each outcome based on a Poisson distribution. We also plotted the weighted cumulative incidence curve for each primary outcome by exposure group. Weighted Cox proportional hazards regression models with robust sandwich variance with stratification by prior OAC treatment were fitted to estimate the hazard ratio (HR) and 95% CI for each outcome associated with rivaroxaban compared with apixaban. We assessed the risk of MACE, MALE using a broader definition, and a composite outcome of MACE and MALE as secondary outcomes.

Secondary analyses

To assess potential effect measure modification, we performed stratified analyses by age (<75 vs. ≥75 years old), sex and clinical presentation of PAD. We investigated whether the risk of each primary outcome varies with prior use of other OACs and with the dose prescribed at cohort entry [standard dose (rivaroxaban 20 mg once daily or apixaban 5 mg twice daily) vs. lower dose]. We also stratified analyses by CHA2DS2-VASc score (<6 vs. ≥6) for the incidence of stroke/TIA/SE, aspirin use in the year before cohort entry and HAS-BLED score at cohort entry (≤4 vs. >4) for the incidence of MB.^18,19^ We used a modified HAS-BLED score without international normalized ratio values as they were not always available. Finally, we assessed the risk for each bleeding site separately (gastro-intestinal bleeding, intracranial haemorrhage, and other bleeding).

Sensitivity analyses

We performed four sensitivity analyses to assess the robustness of the findings. First, we repeated the primary analyses using 15-day and 60-day grace periods for evaluating possible exposure misclassification. Second, we used an intention to treat exposure definition to further investigate potential informative censoring with follow-up limited to 24 months. Third, we ended the study period on 31 January 2020, to avoid any influence of the COVID-19 pandemic. Fourth, we repeated the primary analyses in a restricted population excluding patients with unspecified toe necrosis, non-venous vascular ulcers, unspecified claudication, and aortic bypass surgeries. All analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC).

Results

The study cohort included 6170 new users of rivaroxaban and 9990 new users of apixaban (Figure 1). Table 1 presents the characteristics of the cohort after PS weighting (characteristics before weighting are available in Supplementary material online, Table S2 and characteristics by prior OAC use are available in Supplementary material online, Tables S3 and S4). Most characteristics were similar between exposure groups before weighting, but apixaban users were slightly older and had a higher number of previous hospitalizations. After weighting, both groups were well-balanced with respect to all covariates. The mean follow-up varied from 476 to 486 days for rivaroxaban users and from 502 to 516 days for apixaban users.

Table 2

• presents the results of the primary analyses and weighted cumulative incidence curves for each primary outcome are presented in Figures 2–4. Incidence rates with rivaroxaban compared with apixaban were similar for MALE (weighted incidence 6.7 vs. 5.6 per 1000 person-years; HR: 1,20; 95% CI 0.87–1.65) and stroke/TIA/SE (weighted incidence 24.5 vs. 21.3 per 1000 person-years; HR: 1,15; 95% CI 0.97–1.36). However, the risk of MB was higher with rivaroxaban than with apixaban (weighted incidence 46.1 vs. 29.8 per 1000 person-years; HR: 1,55; 95% CI 1.36–1.77). The HR for MALE was slightly higher among older patients (HR: 1.30; 95% CI (0.84–2.00) than for younger patients (HR: 1.09; 95% CI 0.68–1.77) and for males (HR: 1.30; 95% CI 0.88–1.91) compared with females (HR: 1.06; 95% CI 0.60–1.86) but with wide CIs (see Supplementary material online, Table S5). There was also no effect measure modification on the risk of MALE in other stratified analyses (see Supplementary material online, Tables S6 and S7). Similarly, no difference was observed for the risk of stroke/TIA/SE in stratified analyses (see Supplementary material online, Tables S8–S10). The risk of MALE with the broader definition and the risk of MACE remained unchanged, and there was no difference in the risk of the composite outcome (see Supplementary material online, Table S11). Additionally, the risk of MACE remained consistent when stratified by sex, age and clinical presentation of PAD (see Supplementary material online, Table S12).

Regarding safety, the risk of MB associated with rivaroxaban was not modified in stratified analyses (see Supplementary material online, Tables S13–15). In particular, the risk was higher with both standard dose and low-dose rivaroxaban (see Supplementary material online, Table S15). The risk associated with rivaroxaban remained increased for intracranial haemorrhage, gastro-intestinal bleeding, and other bleeding (see Supplementary material online, Table S16).

Results of sensitivity analyses were also all consistent with those of the primary analyses (see Supplementary material online, Tables S17–20).

Discussion

This population-based cohort study did not show any difference between rivaroxaban and apixaban for the prevention of MALE and stroke/TIA/SE among patients with NVAF and PAD. However, rivaroxaban was associated with a higher risk of MB, compared to apixaban. These results remained consistent across multiple stratified and sensitivity analyses.

Evidence regarding the comparative effectiveness and safety profile of rivaroxaban and apixaban in patients with both NVAF and PAD remains limited in the literature, in particular for MALE, a major outcome in this population. Pivotal PAD trials excluded patients with NVAF and compared rivaroxaban with aspirin. The COMPASS trial showed that low-dose rivaroxaban (2.5 mg twice daily) combined with aspirin reduced MALE and vascular outcomes compared with aspirin alone.^10,20^ This protective effect was also reported with rivaroxaban 5 mg twice daily alone vs. aspirin alone (HR: 0.67; 95% CI: 0.45–1.00).^20^ Similarly, the VOYAGER-PAD trial showed the superiority of low-dose rivaroxaban combined with aspirin compared to aspirin alone in preventing MALE (HR: 0.85; 95% CI: 0.76–0.96).^11^ As a result, the addition of low-dose rivaroxaban to aspirin is strongly recommended in all international guidelines for patients with PAD.^3,21-23^ However, these recommendations specifically exclude patients with NVAF, as they were not included in the aforementioned trials. In our study of patients with NVAF and PAD, which included all dosages of rivaroxaban and apixaban, we observed no difference in the risk of MALE between the two DOACs. There was also no benefit in patients with high-risk limb presentation (rest pain/trophic disorders) compared to patients with intermittent claudication. However, the number of events was relatively low in these subgroups analyses, resulting in wide CIs. As expected, our cohort with NVAF was older and had more comorbidities than patients with PAD alone included in the COMPASS and VOYAGER-PAD trials that compared rivaroxaban + aspirin vs. aspirin alone. Our cohort also differs from VOYAGER that included patients with PAD after recent revascularization whereas we included patients with incident NVAF and history of PAD of any severity. Indeed, patients with PAD in real-world settings tend to have more advanced disease and greater frailty compared to those enrolled in clinical trials.^24,25^

Only one cohort study evaluated the risk of MALE as a secondary outcome with apixaban vs. rivaroxaban in a population with NVAF and type 2 diabetes.^6^ While the overall risk of MALE did not differ between the two DOACs, apixaban was associated with a lower risk of MALE at standard doses (HR: 0.68; 95% CI: 0.47–0.99) compared with rivaroxaban. No difference was observed at lower doses (HR: 0.91; 95% CI: 0.43–1.93). In our study, we did not observe a difference in MALE risk between the two DOACs, even when stratified by dose. This discrepancy may reflect differences in study populations, as our cohort consisted specifically of patients with PAD who likely had a higher baseline vascular risk than the broad population with diabetes.

There is also limited evidence on the effectiveness of rivaroxaban vs. apixaban for the prevention of stroke/SE in this population. One cohort study assessed the risk of stroke/SE in patients aged ≥65 years with NVAF and vascular disease (PAD or coronary artery disease) initiating apixaban or rivaroxaban.^12^ The risk was higher with rivaroxaban compared with apixaban (HR: 1.24; 95% CI: 1.01–1.51). However, <40% of patients had PAD and the analyses were not stratified by PAD status. In our population, no difference was observed between the two DOACs, in line with most studies comparing apixaban and rivaroxaban in patients with NVAF alone.^9,26-31^

Regarding safety, our results are consistent with the aforementioned cohort study in patients with NVAF and vascular disease (PAD or coronary artery disease) that found a greater risk of MB with rivaroxaban compared with apixaban.^12^ However, our study provides significant additional insights by focusing specifically on patients with PAD and including patients across the full clinical spectrum of PAD. Although the comparator differs, our results also align with the COMPASS and VOYAGER trials showing a higher risk of MB with rivaroxaban. We adopted a modified ISTH definition of MB where we classified all bleeding events leading to hospitalization as MB, an approach consistent with the definition used in the COMPASS trial.^10,32^ Notably, the increased bleeding risk observed in our study was consistent across all bleeding sites. This finding contrasts with the results of the COMPASS and VOYAGER-PAD trials, which reported no increased risk of intracranial haemorrhage when comparing low-dose rivaroxaban to aspirin alone.^10,11,32^

This study has several strengths. The CPRD data provided a large, real-world cohort of patients with NVAF and PAD managed in routine clinical practice. Moreover, the linkage to HES and ONS databases reduces potential outcome misclassification. We were able to control for many potential confounders, including lifestyle factors and severity of PAD. Finally, the inclusion of patients with and without prior OAC use ensured generalizability, covering both anticoagulant-naïve and previously treated patients. However, some limitations should be noted. First, residual confounding must be considered given the observational nature of the study. The use of an active comparator and adjustment for an extensive list of confounders through PS weighting helped mitigate confounding by measured factors. However, the potential for bias due to residual confounding cannot be excluded. Second, the identification of patients with PAD relied on diagnostic codes to capture all potentially symptomatic patients with PAD, which enhances generalizability but may introduce misclassification. To address this concern, we conducted a sensitivity analysis restricting the cohort to more specific PAD diagnoses, with results consistent with those of the primary analysis. Third, exposure misclassification is possible as CPRD only records prescriptions issued by GPs. However, since GPs manage most OAC prescriptions for patients with NVAF, this misclassification is likely minimal. Sensitivity analyses using different exposure definitions (15- and 60-day grace periods) produced consistent results, supporting the robustness of our findings. Fourth, the interpretation of potential differences in outcomes beyond the third year of follow-up is limited by the decreasing number of patients remaining at risk. This restricts the ability to draw firm conclusions about the longer-term comparative effectiveness of apixaban and rivaroxaban in this study population. Finally, our cohort was predominantly white, limiting the generalizability of our findings to other ethnic groups.

Overall, in this large population-based cohort of patients with NVAF and concomitant PAD, rivaroxaban showed similar effectiveness compared with apixaban in preventing MALE and stroke/TIA/SE. However, the risk of MB was higher with rivaroxaban, which may help inform the choice of anticoagulant in this population with a complex vascular profile.

Supplementary Material